Niranda Perera

I recently ran the build with the stylecheck and found out a significant number of warnings. We need to do the following - [ ] Enable style check for every...

Currently UCC doesnt support `gatherv`. When it is released, enable gather tests.

Bootstrap UCX processes outside of MPI

1

Parent issue: #575 - [x] Figure out how to create a context out of locally spawned processes (determine what metadata is required to spawn processes) - [x] Get a UCC...

build: UCC integration

8

References: 1. [UCC](https://github.com/openucx/ucc) 2. [torch-ucc](https://github.com/openucx/torch-ucc) 3. [torch-ucc fb research](https://github.com/facebookresearch/torch_ucc) Note: UCX requires 1.11

Currently we have two UCP context objects. 1. From our previous UCX integration 2. New UCC context init also uses MPI and creates a new UCP context underneath These need...

Currently `cylon::DataType` mimics [`arrow::DataType`](https://github.com/apache/arrow/blob/master/cpp/src/arrow/type.h) which is a class-based type declaration. But we are not using the type classes that much. [`cudf::DataType`](https://docs.rapids.ai/api/libcudf/stable/types_8hpp_source.html) follows an enum-based type declaration, and I believe this...

Extend sorting and may be reuse this for distributed quantile operation as well. The algorithm is discussed here https://docs.google.com/document/d/1nicrPuDepPLYSYsvdmkE56TPXO4C4azDUcl15u08i6s/edit# GPU distributed sorting https://github.com/cylondata/cylon/pull/510

Only keep 1. Conda arrow build 2. Pip arrow build 2. Externally compiled arrow build Deprecate 1. Local arrow build (which the current default option)